Thank you for trying the Library Explorer :) Very interesting talk! Personally, I don't think anything there is a testament for abandoning classification trees

Thank you for trying the Library Explorer :) Very interesting talk! Personally, I don't think anything there is a testament for abandoning classification trees entirely; just "common pitfalls" to avoid when building new ones. The Library Explorer is very purposefully classification-system agnostic, so if a better classification system comes along, we can import it and switch to it! And unlike a physical library, we don't have any re-shelving costs :) One of the core design features I was aiming for with Library Explorer was that the user should never be navigating the "hierarchy;" they should be navigating the books. Forcing the user to move through categories forces them to try to understand the hierarchy, which can at times be not super user-friendly. Note that the Yahoo/Google examples in the video do just that; the user picks a class, and then sees websites. By showing them the books directly, the user "deduces" the classification tree (lots of books about physics? I must be in the physics section); the classification labels aren't really necessary for exploration. In general, I think there are trade-offs. A classification tree is a model with flaws (just like all models). But its core benefits are (1) it's a tree; so each node has semantic siblings, parents, and children; (2) each node in the tree has a finite number of child nodes (and usually <30 child nodes; which makes it ideal for human traversal); (3) each book can be uniquely identified by the path from the root. (3) Was very important for physical librarians (classification systems were sort of like a search index for librarians), and was the cause of some of the issues described by Shirky. But online, we don't need (3), so we can kind of throw it away. The tree can be degenerate; books can (and do!) appear in multiple nodes of the classification tree, because, unlike in a physical collection, in a digital collection, books are disjointed from the classification tree itself. Tag-clouds are _definitely_ more flexible (as Shirky said, classification trees are a _restriction_ on the tag-cloud model), but come at a cost of being harder to navigate (as flexibility usually does) and get a "big picture" idea of. I haven't seen a good example of a UI that lets you navigate tags in a way that doesn't just let you travel from one node to a neighbouring node (please post if folks have one!). Trees allow you to travel up (parent node), down (child node), and left/right (sibling nodes). Graphs you're limited to just traveling to adjacent nodes. You could try to algorithmically _deduce_ parent/child relationships though (that would be an interesting thing to try! I'm sure algorithms exist that do this). Ways of browsing graphs is definitely an interesting problem space! Open Library does also have tags (called "Subjects"), so finding a way to make them more user-browsing friendly would be great :)